How a Denver ed-tech company standardized AI across four PMs without slipping a release

The Head of Product had spent three years pulling the org out of a feature-request culture. The roadmap tied back to three growth areas, and everything else got cut. AI adoption was equally real: Claude ran in every PM's day, prototypes went through Lovable before design touched them, and Gemini sat inside every Google doc.

The problem was never adoption. It was what came after. Four PMs had four private stacks, four sets of prompts, and four versions of what good looked like. The Head of Product measured it before rollout: each PM was spending four to six hours a week just evaluating new tools, and cross-PM review of AI-assisted work had quietly stopped happening.

The team consolidated onto Second Axis over one sprint boundary, mid-quarter, with a hard rule: if the migration threatened the semester release, it stopped. It never came close. Existing prompts and working patterns were mapped into shared workflows during onboarding, so the first week felt like a faster version of the old routine rather than a new tool.

The part nobody expected was the feedback loop. For the first two weeks, PMs thumbs-downed everything the morning brief got wrong: flagged noise, missed decisions, wrong priorities. By week three the corrections stopped, because the brief had learned what this team considers worth interrupting a morning for. What used to be four private routines became one system the whole team had trained.